Kimia Shop
FAEN

Ready workload templates

Three steps: choose the workload, pick the server model, and take the one of the three tiers «Economy», «Medium», «Advanced» that fits your need. Each tier is built by deterministic rules from the official catalog attributes and evaluated «error-free» by the QuickSpecs compatibility engine; before ordering you can swap, add or remove any part.

Each tier is built by deterministic rules from catalog attributes (processor cores/frequency/TDP, memory channels, drive capacity and type, controller class, network speed, power-supply wattage) and is published only when the QuickSpecs compatibility engine reports no error and the tier is not smaller than the one below it; when no combination is possible the reason is written there. Other mandatory parts (heat sinks, cables, batteries, …) are suggested by the compatibility engine. No prices are shown.

Workload

VirtualizationHost for many virtual machines: many cores, many memory channels and mixed-use endurance SSDs.DatabaseTransactional database: high frequency, lots of memory and fast storage (NVMe/SSD) with mixed-use endurance.File server and storageFile / archive / backup with high capacity: LFF chassis, high-capacity hard drives and a RAID controller.Virtual desktop (VDI)Virtual desktop host: high-frequency processors, lots of memory, graphics accelerators (when available) and SSDs.AI and GPUTraining / inference: the most GPU memory in the catalog, system memory of at least twice the GPU memory, and NVMe.Web and applicationWeb and application server: balanced processors, medium memory, read-oriented SSDs and a fast network.

Server model for «AI and GPU»

Dell PowerEdge R760No GPU in the catalogIntel Xeon Scalable 4 5 ( Xeon Max)This model's catalog has no GPU with a specified graphics memory; an AI/GPU template is not built.
HPE ProLiant DL380 Gen12Intel Xeon 6 ( E-Core P-Core)
HPE ProLiant DL580 Gen12No GPU in the catalogXeon 6 ( P-Core)This model's catalog has no GPU with a specified graphics memory; an AI/GPU template is not built.
HPE ProLiant DL380 Gen11Intel Xeon ScalableHPE ProLiant DL380 Gen10 PlusIntel Xeon ScalableHPE ProLiant DL380 Gen10Xeon Scalable (Skylake / Cascade Lake)HPE ProLiant DL580 Gen10Intel Xeon Scalable (82xx/62xx/52xx) (81xx/61xx/51xx)Lenovo ThinkSystem SR650 V3Intel Xeon ScalableHPE ProLiant DL380 Gen9Xeon E5-2600 v4
HPE ProLiant DL580 Gen9No GPU in the catalogXeon E7-4800/8800 v4This model's catalog has no GPU with a specified graphics memory; an AI/GPU template is not built.
HPE ProLiant DL380e Gen8No GPU in the catalogIntel Xeon E5-2400 E5-2400 v2This model's catalog has no GPU with a specified graphics memory; an AI/GPU template is not built.
HPE ProLiant DL380p Gen8E5-2600 E5-2600 v2HPE ProLiant DL580 Gen8Xeon E7-4800/8800 v2

Three «AI and GPU» tiers onHPE ProLiant DL380 Gen10 Plus

Economy

AI and GPU

0 warning(s) · loading the compatibility engine…
Processor
1 × Xeon Silver 4314
16 cores in total · 2.4 GHz · 135 W
Memory
128 GB
8 modules × 16 GB
Storage
2 drives · 3.2 TB raw
1.6 TB usable (RAID 1)
Network
2 ports · 20 Gb
Fastest port 10 Gb
Power supply
2 × 800 W
Power-supply redundancy
GPU
1 pcs · 16 GB
Why this tier?
  • Processor: 1 × Xeon Silver 4314 — 16 cores in total (16 cores, 2.4 GHz, 135 W per processor). The largest model usable in this configuration has 40 cores and the rule of this tier asks for at least 35% of that.
  • Memory: 8 modules of 16 GB = 128 GB. This platform has 8 memory channels per processor; 8 modules per processor means 1 module(s) per channel and all channels are filled equally (the QuickSpecs memory-balance recommendation).
  • Storage: 2 NVMe drives of 1.6 TB (endurance MU) = 3.2 TB raw; 1.6 TB usable with RAID 1 (standard RAID formula, no overhead).
  • Network: 1 adapter(s) — 2 ports and 20 Gb nominal capacity in total (fastest port 10 Gb).
  • GPU: 1 with 16 GB of graphics memory in total; system memory of at least twice the GPU memory is assumed (QuickSpecs recommendation).
  • Power supply: 2 × 800 W (power-supply redundancy).

Selection rule of this tier

  • The chassis with the fewest drive bays that still fits the drives of this tier (SFF, 2 drives)
  • 1 processor(s) — the smallest model that has at least this share of the cores of the largest catalog model (35%), maximum TDP 300 W
  • Memory: about 128 GB per processor; number of modules = memory channels (8 channels per processor) × number of processors, one module per channel (two when needed) so that the channels are filled evenly
  • Storage: NVMe (if unavailable: SSD), endurance MU, 2 drives with a capacity at the 40th percentile of the available capacity list
  • Network: an adapter with a total capacity close to 20 Gb
  • Power supply: the smallest AC model with at least 1 times the estimated peak consumption (processor TDP + GPU + 200 W), two identical supplies
  • GPU: 1, the least graphics memory (at least 16 GB)
Open in configuratorTechnical report

Medium

AI and GPU

0 warning(s) · loading the compatibility engine…
Processor
2 × Xeon Gold 6342
48 cores in total · 2.8 GHz · 230 W
Memory
512 GB
16 modules × 32 GB
Storage
4 drives · 12.8 TB raw
9.6 TB usable (RAID 5)
Network
2 ports · 50 Gb
Fastest port 25 Gb
Power supply
2 × 800 W
Power-supply redundancy
GPU
2 pcs · 96 GB
Why this tier?
  • Processor: 2 × Xeon Gold 6342 — 48 cores in total (24 cores, 2.8 GHz, 230 W per processor). The largest model usable in this configuration has 40 cores and the rule of this tier asks for at least 60% of that.
  • Memory: 16 modules of 32 GB = 512 GB. This platform has 8 memory channels per processor; 8 modules per processor means 1 module(s) per channel and all channels are filled equally (the QuickSpecs memory-balance recommendation).
  • Storage: 4 NVMe drives of 3.2 TB (endurance MU) = 12.8 TB raw; 9.6 TB usable with RAID 5 (standard RAID formula, no overhead).
  • Network: 1 adapter(s) — 2 ports and 50 Gb nominal capacity in total (fastest port 25 Gb).
  • GPU: 2 with 96 GB of graphics memory in total; system memory of at least twice the GPU memory is assumed (QuickSpecs recommendation).
  • Power supply: 2 × 800 W (power-supply redundancy).

Selection rule of this tier

  • The chassis with the fewest drive bays that still fits the drives of this tier (SFF, 4 drives)
  • 2 processor(s) — the smallest model that has at least this share of the cores of the largest catalog model (60%), maximum TDP 300 W
  • Memory: about 256 GB per processor; number of modules = memory channels (8 channels per processor) × number of processors, one module per channel (two when needed) so that the channels are filled evenly
  • Storage: NVMe (if unavailable: SSD), endurance MU, 4 drives with a capacity at the 60th percentile of the available capacity list
  • Network: an adapter with a total capacity close to 50 Gb
  • Power supply: the smallest AC model with at least 1.2 times the estimated peak consumption (processor TDP + GPU + 200 W), two identical supplies
  • GPU: 2, medium graphics memory (at least 16 GB)
Open in configuratorTechnical report

Advanced

AI and GPU

0 warning(s) · loading the compatibility engine…
Processor
2 × Xeon Platinum 8380
80 cores in total · 2.3 GHz · 270 W
Memory
1024 GB
16 modules × 64 GB
Storage
8 drives · 51.2 TB raw
44.8 TB usable (RAID 5)
Network
8 ports · 200 Gb
Fastest port 25 Gb
Power supply
2 × 1800 W
Power-supply redundancy
GPU
3 pcs · 192 GB
Why this tier?
  • Processor: 2 × Xeon Platinum 8380 — 80 cores in total (40 cores, 2.3 GHz, 270 W per processor). The largest model usable in this configuration has 40 cores and the rule of this tier asks for at least 100% of that.
  • Memory: 16 modules of 64 GB = 1024 GB. This platform has 8 memory channels per processor; 8 modules per processor means 1 module(s) per channel and all channels are filled equally (the QuickSpecs memory-balance recommendation).
  • Storage: 8 NVMe drives of 6.4 TB (endurance MU) = 51.2 TB raw; 44.8 TB usable with RAID 5 (standard RAID formula, no overhead).
  • Network: 2 adapter(s) — 8 ports and 200 Gb nominal capacity in total (fastest port 25 Gb).
  • GPU: 3 with 192 GB of graphics memory in total; system memory of at least twice the GPU memory is assumed (QuickSpecs recommendation).
  • Power supply: 2 × 1800 W (power-supply redundancy).

Selection rule of this tier

  • The chassis with the fewest drive bays that still fits the drives of this tier (SFF, 8 drives)
  • 2 processor(s) — the smallest model that has at least this share of the cores of the largest catalog model (100%), maximum TDP 300 W
  • Memory: about 512 GB per processor; number of modules = memory channels (8 channels per processor) × number of processors, one module per channel (two when needed) so that the channels are filled evenly
  • Storage: NVMe (if unavailable: SSD), endurance MU, 8 drives with a capacity at the 90th percentile of the available capacity list
  • Network: an adapter with a total capacity close to 100 Gb (2 adapters)
  • Power supply: the smallest AC model with at least 1.4 times the estimated peak consumption (processor TDP + GPU + 200 W), two identical supplies
  • GPU: 4, the most graphics memory
Open in configuratorTechnical report
Loading the catalog for customization and comparison…